Evaluating Commercial Electrical Neuromodulation Devices with Low-Cost Neural Phantoms

Abstract

Non-invasive transcranial electrical stimulation is a category of neuromodulation techniques used for various disorders. Although medically approved devices exist, the variety of consumer electrical stimulation devices is increasing. Because clinical trials and animal tests are costly and risky, using a brain phantom can provide preliminary experimental validation. However, existing brain phantoms are often costly or require excessive preparation time, precluding their use for rapid, real-time optimization of stimulation settings. A limitation of direct electric fields in a phantom is the lack of 3D spatial resolution. Using well-researched modalities such as transcranial direct current stimulation (tDCS) and newer modalities such as amplitude-modulated transcranial pulsed-current stimulation (am-tPCS), a range of materials was tested for use as electrical phantoms. Based on cost, preparation time, and efficiency, ground beef and agar gel with a 10% salt mix were selected. The measured values for the total dosages were 0.55 W-s for am-tPCS and 0.91 W-s for tDCS. Due to a low gain on the recording electrodes, the signal efficiency measured against the power delivered was 4.2% for tDCS and 3.1% for am-tPCS. Issues included electrodes shifting in the soft material and the low sensitivity of the recording electrodes. Despite these issues, the effective combination of the phantom and recording methodologies can enable low costs and the rapid testing, experimentation, and verification of consumer neuromodulation devices in three dimensions. Additionally, the efficiency factors (EFs) between the observed dosage and the delivered dosage could streamline the comparison of experimental configurations. As demonstrated by comparing two types of electrical neuromodulation devices across the 3D space of a phantom, EFs can be used in conjunction with a cost-effective, time-expedient phantom to rapidly iterate and optimize stimulation parameters.

1. Introduction

1.1. Overview

Non-invasive transcranial electrical neuromodulation is a category of techniques for treating various neurological disorders [1,2,3,4,5,6,7,8]. A variety of electrical neuromodulation devices are sold as non-medical devices, which differ in modality and quality [7,9]. However, these often involve significant financial costs and ethical considerations. Obtaining and maintaining living tissues, along with the potential need for regulatory approval due to working with living subjects, can substantially increase the complexity and cost of experimental setups. A brain phantom, or material simulating the brain’s electrical properties, can be used to acquire experimental results without animal or human testing [10,11,12]. Material selection, electrode placement, and costs in money and time are all factors in experimental design [13,14,15,16]. A low-cost neural phantom could be used to reliably measure and verify the power delivered by a device and its mechanism of action. Prior phantoms, however, can often take days of preparation or require specialized, expensive materials [13,14,15,16]. Separate performance metrics have also been used to compare phantoms, such as the electric field. Such limitations preclude rapid, iterative adjustments of stimulation parameters. Calculating the electric field often assumes a uniform distribution across the phantom material, which may not be the case with heterogenous materials and unconventional stimulation modalities. By determining a cost-effective, easily prepared brain phantom material and proposing a common performance metric, electric neuromodulation parameters could be rapidly tested and compared. By proposing a common performance metric, three-dimensional electric field measurement technique, and material preparation comparison, several low-cost phantoms and measurement parameters were explored to validate two consumer electrical neuromodulation devices.

1.2. Background

Non-invasive brain stimulation for therapeutic purposes, or neuromodulation, has been consistently used to manage various neurological and psychiatric disorders [1,3]. Non-invasive neuromodulation, unlike invasive deep-brain stimulation with implanted devices, does not require complex surgery [6,17]. Applied electric currents have been used for non-invasive neuromodulation since ancient Egypt [18]. Contemporary techniques employed include transcranial direct current (DC) stimulation (tDCS) and amplitude-modulated transcranial pulsed-current stimulation (am-tPCS) [1,4,19]. Together with other electrical stimulation methods, tDCS has been extensively explored in research and medical literature for decades [20]. Conversely, the novelty of am-tPCS has precluded widespread understanding of its possible efficacy and mechanisms of action.

Commercial tDCS devices apply a low-amperage direct current between a positive anode and a negative cathode [8,21,22]. Larger sponge pads are used to reduce the risk of burns, as well as a lower current density [9]. Discrete patterns of electrode placement, stimulation currents, and stimulation time are referred to as “montages”. Individual montages were reported to stimulate specific brain states (for example, heightening awareness and response time). Commercial tDCS devices stimulate for approximately 20–30 min to avoid delivering potentially hazardous amounts of electrical power to the brain [6]. The dominant theory for the tDCS mechanism of action concerns lowering the resting potential of a brain region by saturating it with current, thereby facilitating action potential propagation in that region [20,23]. The opposite case is the potential inhibition of action potential propagation while temporarily raising regional resting potential. This model has been challenged but remains widely discussed [4,19,24].

Conversely, am-tPCS combines other montages of electrical stimulation. In addition to tDCS, other types of transcranial electrical stimulation include transcranial alternating current (AC) stimulation (tACS), transcranial pulsed-current stimulation (tPCS), and transcranial random noise stimulation (tRNS) [25,26]. Whereas tACS uses a varying sinusoidal electrical stimulus pattern [27], tPCS uses binary pulses of direct current, as opposed to the single, constant level used in tDCS [5,26,28]. Whereas tRNS uses alternating current sinusoidal stimulations with randomized frequencies and amplitudes, am-tPCS combines tPCS and tRNS, using pulses of DC and amplitude-controlled AC random noise [1]. Every 2.5 s, the polarity of the device changes; the manual claims it can entrain endogenous oscillations [29]. The mechanism of action was hypothesized to combine aspects of tPCS and tRNS, ensuring that neurons acclimate less easily to a constant DC [28]. Live test data on am-tPCS are limited.

The stimulation modalities also generate different localized polarizations in the electric field. For example, the sinusoid shape of tACS waveforms may travel at different speeds through different media [27]. Even time-varying measurement at two points may not capture the entire profile of electrical activity. A finite element-based method, even involving relatively direct arithmetic, can offer a potential way to estimate local electric field activity. Prior approaches have involved complex software simulations and multi-electrode recording systems [30,31,32]. Simplifying the requirements for electric field measurement and characterization would streamline high-resolution electric field measurements.

Animal and clinical human trials can be costly and potentially dangerous [9]. Electrical stimulation, even those that are non-invasive, can endanger a living being [33]. From atrial fibrillation to burns, the possible mechanisms of harm are well documented [6,9]. Animal and human studies also have a greater cost in time, adding another potential variable to this study. Tissues donated by a living organism have a limited lifespan without specialized care. Although cultured cells and organoids can be used as a substitute, these also require specialized care and maintenance [2]. Conversely, non-biological phantoms and simulated tissue have been used to accurately model organic tissues [12,13,14,16]. Brain phantoms have been produced from a range of materials, including solids and liquids [14,15]. To accurately model the effects of applied electrical stimulation, the brain phantom must contain electrolytic compounds comparable to the original organ [10]. Although computer simulations have been used for estimates, phantom tissue enables experimental results without requiring human or animal data [10,14,34,35]. Many materials have been used as brain phantoms, with variable costs and preparation times. Different studies use different metrics, making direct comparisons more difficult [12,13,14,16]. Thus, directly comparing different electrical stimulation modalities on the same phantom becomes challenging.

A wide range of electrical stimulation devices are marketed and sold as non-medical therapeutic devices [33,36]. The efficacy of these “off-label” devices is controversial in the research literature, even compared to medically approved devices [9]. The range of manufacturers and relatively low cost of electrical stimulation devices have resulted in a range of products, from hobbyist kits to high-quality devices [6]. The potential reliability and safety of these devices could be more easily evaluated with a phantom [13,16]. Different circumstances and experimental research often require the rapid, iterative optimization of stimulation for specific circumstances. Having a consistent performance metric, and readily prepared phantom and measurement method could streamline the optimization of the stimulation parameters. A low-cost, low-complexity phantom could enable the affordable and economical testing of consumer electrical neuromodulation devices, without the risk of hurting animal or human subjects, while economizing on time and money.

2. Materials and Methods

2.1. Overview

A low-cost phantom enables the evaluation of electrical stimulation devices without risk to animals or humans. However, many phantoms have been reported in the literature. Phantom brains include ground beef, agar gel mixed with salt, and phosphate-buffered saline (PBS) [13,16]. Cost and preparation time are primary considerations because experimental technologies often require rapid iteration at low cost. Thus, finding accurate phantoms that are low cost and need minimal preparation time can reduce the cost of individual experiments. Following this, a performance metric and method based on sensor sensitivity was proposed.

2.2. Phantoms

Three materials were examined for use as brain phantoms: ground beef, agar gel mixed with salt, and PBS [10,12,13]. Each material has been used in at least one peer-reviewed study, including those with successful subsequent animal and human trials. Ground beef, purchased directly from a supermarket, has been used as a phantom in several studies. Agar gel, or ballistic gel, was mixed with salt in different ratios: 1%, 5%, and 10%. PBS was mixed with boiling water and used to model a brain. The preparation times and costs (in US dollars) were compared. As shown in

Table 1. Comparison between prior phantom models and the current work.

	Agar Gel w/Salt	PBS	Beef	This
Reference	Owda (2021) [12]	Guido (2019) [10]	Guido (2020) [11]	N/A
Type	-Electric Stim	-DBS Testing	-RF Heating	-Electric Stim
Time (h)	24	1	<1	1–24
Cost (USD)	USD 25	USD 60	USD 10	USD 10–25
Measure	Voltage (V)	Voltage (V)	Thermal, Temp	Voltage (V), EF
Limits	-Long prep time	-Costly	-Different domain	-Purpose-built
Benefits	-Customizable	-Chemically close	-Cheap, fast setup	-Cheap, fast setup

, the materials with the lowest cost and preparation time were used.

Each material was prepared according to methods reported in prior research. Agar gel was measured with respect to the weight of distilled water in the container, and salt was added in proportions of 1%, 5%, and 10% of the total agar gel mass [12]. In the case of the 10% mixture, the agar gel was prepared by heating 0.95 L of distilled water, used to dissolve with 108 g salt. A total of 59.2 mL of agar gel powder was mixed in. The mixture was stirred and boiled for 5 min on a stovetop at 100 °C. It was removed from the stove, left to cool for 5 min, and poured into containers. It was refrigerated overnight. It was warmed to room temperature (25 °C) before experimentation [12]. While ground beef has variable composition, the same brand was purchased from the same supermarket for all tests. Prior studies with ground beef reported that despite its variable composition, it was sufficiently homogenous in its electrical properties [11,37]. PBS was boiled in distilled water. Agar gel was prepared in batches 24–48 h before the experiment and refrigerated. Beef was purchased the day of the experiment, and emptied into half the container’s volume. PBS was prepared shortly before experimental usage. Containers with the same volume (1280 mL) were used in all cases. Each sample in a container was reused until all measurements of its material were completed.

The conductance and impedance of each material was documented elsewhere. The brain itself has a conductivity of 0.26 S/m for gray matter and 0.17 S/m for white matter, but the skin resistance must be accounted for with non-invasive neuromodulation [38]. PBS mimics the ion concentrations and pH inside the human body, with an impedance of 33 kΩ at 0.5 Hz. It has been used repeatedly in stimulation phantom models [10,39]. Meat products, such as ground beef, have an impedance and conductivity dependent on power dissipated [37]. The reported electrical conductivity of meat ranged from 0.01–10 S/m. At room temperature (25 °C), the electrical conductivity of ground beef was reported as approximately 1.6 S/m [37]. Others have reported success using ground beef as a brain phantom in regard to both its electric and thermal properties [11]. Agar gel mixed with salt has long been used as an electrical brain phantom, although the electrical properties depend on the salt [40]. The measured electrical resistance was at 200–300 Ω. The addition of even 10% salt (NaCl) was reported to drop impedance by a factor of 14 [12]. Although skin conductance was not factored into these prior studies, published research reported that all three materials should sufficiently act as brain phantoms for room temperature experiments conducted with low-powered neuromodulation devices. While a 3D measurement technique could detect electric fields at a high resolution across the phantom, different stimulation devices would generate different profiles, even at a low power stimulation.

2.3. Device Characteristics

Two devices were directly compared with each other. The first was a commercially available BrainDriver v2.1 tDCS device (TheBrainDriver, Chicago, IL, USA). The second device was an Axis am-tPCS device (U: The Mind Company, Columbus, OH, USA). The stimulation parameters in the BrainDriver were user-selected, ranging from 0.5 to 2 mA in current and with a stimulation duration between 20 and 30 min [41]. The BrainDriver had a constant DC voltage of 9 V. The device had a single anode and single cathode, each with a sponge electrode 2 cm in diameter, contained in an appropriately sized rubber sheath. Parameters for neuromodulation were based on prior work, with a current of 2 mA applied for 20 min [9,42]. The sponges were soaked in saline solution prior to active use.

The Axis device’s two electrode pads were designed differently [29]. Each electrode pad was a 3D printed plastic block with one side covered by a 16.1 cm² pad of copper tacks facing inwards. A scattered array was used to preclude current saturation. A constant voltage of 5 V was applied with a current of 4 mA. Each pulse comprised 0.9 ms on and 0.1 ms off. The anode and cathode were altered every 2.5 s to entrain local connections. The stimulation program ran for a fixed duration of 15 min. The manufacturer reported that the Axis device uses less total power than comparable devices [29]. Electrodes from the Axis device were coated in conductive saline solution. For both devices, electrodes were positioned in fixed locations, secured with electrical tape if necessary. Although the stimulation time and applied power from each device differed, the phantoms were tested with a complete, manufacturer-recommended dosage.

2.4. Calculations

The total power for each device was calculated using Equation (1). The voltage V is multiplied by the current I.

$$P=V\ast I$$

(1)

The total amount of applied electricity for a particular montage and stimulation period is called the “dosage”. As shown in Equation (2), the dosage D is calculated as a function of the power P and stimulation duration t.

$$D=P\ast t$$

(2)

In both metrics, the tDCS device delivers more power into the brain. Although the electrical impedance of human skin has a broad range, the Thevenin equivalent circuit was calculated and is displayed in Figure 1 [43]. In both devices, the voltage V was applied with an adaptive resistance R1 to apply a constant current. The constant current faced the dynamic impedance of human skin in the form of R2. Due to the dynamic nature of impedance, the dosage was more important than raw voltage or current values for electrical stimulation [2,25].

The dosages delivered by both devices were calculated using the equations and model, and are provided in

Table 2. Calculated dosages for tDCS and am-tPCS devices.

	Current (A)	Voltage (V)	Power (W)	Time (min)	Time (s)	Dose (W-s)
am-tPCS	0.004	5	0.02	15	900	18
tDCS	0.002	9	0.018	20	1200	21.6

.

The total tDCS dosage of the BrainDriver was slightly higher than that of the am-tPCS Axis. A device delivering a higher dosage may not equate to a more substantial effect. Experimental observation of each device delivering the specified dosage could confirm manufacturer claims for each non-medical device.

2.5. Measurement Systems

A specialized 10-electrode electrophysiological recording system was designed to measure voltage and current. Figure 2 shows the circuit diagram used in the electrode probe measurements, including reference and ground, denoted by red-outlined yellow squares on the breadboard. A total of nine active electrode channels were supported simultaneously, one in each of the nine cubic volumes. Measurement electrodes were modified from multimeter probes, terminating in thin metal conical tips, 0.7mm in diameter.

As shown in Figure 2, the device used a Teensy 3.2 microcontroller (PJRC, Portland, OR, USA) to acquire the signals at a button press. With each button press, each electrode was sampled at 100 Hz. The average measurement of each electrode was exported to a computer connected via a USB cable. Each phantom was prepared in a 1280 mL (~43.3 fl oz) plastic container. Each electrode corresponded to a specific cubic volume in the container of the phantom material. As shown in Figure 3, the volumes corresponding to the pathway between stimulation electrodes were averaged together. The cubic volumes were denoted by a letter for position horizontally and number for vertical depth (e.g., A0, B2, C1, etc.). A total of nine cubic volumes were recorded in each experiment: A0, A1, A2, B0, B1, B2, C0, C1, and C2.

If the measurement device could not be reliably used after three attempts, a redundant measure was used. A common multimeter was set to DC voltage measurement, and the probes were then inserted into locations and depths corresponding to each cubic measurement. A total of three measurements per cubic volume were conducted and then repeated with the probe positions switched. The schematic of the system is further detailed in Figure 4. The phantom material had electrode probes inserted into the sections detailed in Figure 3. The acquisition board, detailed in Figure 2, connected the probes with a power supply via USB cable and the data to an external device, such as a PC or laptop. The stimulation device was operated independently, with stimulation electrodes positioned after recording electrodes were inserted.

The average values were used similarly to the measurement system. Because the phantom materials were flexible and soft, constant shifting and movement meant that the electrodes may not have been in continuous contact with the material, potentially affecting the observed values.

2.6. Measurement Steps

The substrate was prepared and left to acclimate to room temperature for 1 h. Stimulation electrodes were positioned in two separate configurations: parallel and perpendicular. Electrodes were 10 cm apart in each case. Before placement, measurement electrodes were calibrated with a common DC voltage source, a fresh AA battery at 1.5 V. The distance was factored into the electrode placement to calculate the voltage gradient [4,22]. The absolute distance in the parallel case was higher, allowing more space for charge to dissipate into the material. The perpendicular case had a shorter distance between the stimulation electrodes. The measurement and stimulation devices were coated in conductive electrophysiological saline gel to reduce impedance. The voltage measured per distance was calculated for each and averaged. This measurement sequence was repeated twice per sample. If a material proved too pliable, the electrodes could shift during measurement.

The stimulation electrodes were positioned on the surface for each device in both parallel and perpendicular configurations, as the electric field was perpendicular to the measurement [14,30]. The deeper layers were farther from the stimulation, so the measurements from each layer were averaged together to consolidate and filter readings, as in Equation (3).

$$V_{avg}={\displaystyle \frac{1}{3}}{(V_{A2}+V_{B2}+V}_{C2})+{\displaystyle \frac{1}{3}}{(V_{A1}+V_{B1}+V}_{C1})+{\displaystyle \frac{1}{3}}{(V_{A0}+V_{B0}+V}_{C0})$$

(3)

Averaging the voltage recordings from each cubic volume enabled robust, residual changes from electrical stimulation, as opposed to momentary shifts due to stimulation modality or noise [14,30]. For both perpendicular and parallel electrode placements, the average electric field $V_{,avg}$ was calculated using Equation (3). Assuming both were the legs of a right-angled triangle, Pythagorean theorem was used to calculate the net average voltage $V_{net,avg}$ in Equation (4) [14,30]. The measured $V_{net,avg}$ was then divided by the size of each cubic region to calculate the electric field.

$$V_{net,avg}=\sqrt{{\left(V_{net,parallel}\right)}^2+{\left(V_{net,perpendicular}\right)}^2}$$

(4)

The measured $V_{net,avg}$ would reflect the measured efficiency. Reliable contact was key to low impedance, and thus high sensitivity, for the measurement electrodes. An efficiency factor EF was calculated for each device to measure potential power loss.

$$EF(\%)=D_{obs}/D_{exp}$$

(5)

As shown in Equation (5), the EF was the ratio of the observed dosage D_obs over the expected dosage D_exp. The observed power was based on experimental measurements. The expected power was based on values from Equation (2). The combined percentage represented the fraction of power captured. If a material’s efficiency factor was below 0.5%, it was excluded from future tests. Material cost and preparation time were also criteria for exclusion. The experiments were conducted at a lab maintained at 25 °C, at an indoor relative humidity of 50%. Each batch of materials was prepared in advance of the experiment. Hourly temperature and electrical measurements, without equipment, ensured that the samples were reliable at room temperature for at least 2 h. The EF depended on the reliability of the dosage measurement, including electrode geometry, electrode material, device configuration, and experimental design. As such, the EF could serve as a comparison method between phantom experiments for non-invasive neuromodulation devices.

2.7. Hypotheses

For the ground beef, PBS, and gel mixes, it was hypothesized that similar activity would be observed across each. Although the materials possessed different concentrations of electrolytes, all had been reported in prior peer-reviewed literature. However, electrode placement could introduce significant inefficiency to the system. If external factors (such as ambient temperature) excessively thawed the gel, the efficiency factor would be too low. If the measurement system was insufficiently sensitive or possessed too low a gain, the efficiency factor would be low. If manufacturers’ specifications were accurate, each device would have a voltage gradient between 3.8 and 33.3 mV/mm [29,41].

3. Results

3.1. Overview

The first tests focused on material cost and preparation time. Following this, a full battery of tests was conducted with the remaining materials. The experimentally observed values were used to calculate the efficiency factor for each device and material. The final power, dosages, and efficiency factors between the two devices were calculated and compared to manufacturer specifications.

3.2. Phantom Costs

The materials varied in cost and preparation time.

As shown in Figure 5, the ground beef had the shortest preparation time, being simply left to thaw or heated in a common microwave. The agar gel took approximately 24 h; a day to mix and prepare. The PBS required an hour to boil with water. A consistent volume container was used to compare all materials. The costs per unit were calculated for each material. The highest cost was for PBS, at USD 60. As shown in Figure 6, the second highest cost was for agar gel and salt, at USD 25. The lowest price was for commercial ground beef, at USD 10.

Despite its short preparation time, the high cost of PBS prompted its exclusion from further testing. Of the three salt and gel mixtures, the greatest consistent voltage gradient was measured with a 10% salt content. The 1% and 5% salt mixtures were excluded to eliminate redundant measurements and streamline preparation time. Further tests were conducted with the ground beef and the 10% salt–gel mixture.

3.3. Experimental Measurements

As shown in Figure 6 and Figure 7, electrical measurements were conducted with gel and beef in parallel and perpendicular configurations. When stimulating with tDCS electrodes in the perpendicular position, the measured average voltage gradient was 26.25 ± 8.8 mV/mm in the ground beef. With the am-tPCS stimulation in the perpendicular position, the average voltage measured 13 ± 2.6 mV/mm with the 10% salt–gel mix.

For parallel electrodes with tDCS, the average voltage was 31.5 ± 10.5 mV/mm with the 10% salt–gel mix.

For the am-tPCS electrodes in the parallel configuration, the highest average voltage was 14.25 ± 4.6 mV/mm with the 10% salt–gel mix. The gel and salt mix, after the removal of the electrodes and recording system, is illustrated in Figure 8.

As shown in Figure 9, the highest average voltage gradient from the combined model was 38.75 ± 1.0 mV/mm with tDCS in beef.

3.4. Efficiency Factors

As shown in

Table 3. Efficiency factors for power and dosage measurements.

	Current (A)	Vg (mV/mm)	Dose Observed (W-s)	Dose Expected (W-s)	EF (%)
am-tPCS	0.004	15	0.55	18.00	0.031
tDCS	0.002	38	0.91	21.60	0.042

, the experimentally observed values were used to calculate the power. The measured voltage gradients used were 37.8 ± 0.9 mV/mm for tDCS and 15.35 ± 3.9 mV/mm for am-tPCS.

Based on experimental observations, the estimated dosages were 0.55 W-s for am-tPCS and 0.91 W-s for tDCS. The efficiency factors measured were 4.2% for tDCS and 3.1% for am-tPCS.

4. Discussion

4.1. Overview

A phantom brain facilitates the safe, ethical testing of neuromodulation devices. A low-cost, time-expedient phantom brain with a consistent performance metric enables the comparison and systematic experimentation of neuromodulation parameters. The effects of tDCS on the human brain are well documented, whereas those of am-tPCS are less so. Developing a reliable phantom enables experimentation with new neuromodulation methods and performance validation of commercial devices [10,16]. The measured voltage gradients were comparable with manufacturer specifications [29,41]. Using low-cost materials, the experimental techniques developed here could reduce development costs for new electrical stimulation. Although the am-tPCS device was not tested on a living being, it performed comparably to the commercial tDCS device. The total dosages delivered remain safe for human use [9]. Although the recording system had an efficiency of 4.2% for tDCS and 3.1% for am-tPCS, it remained capable of measuring the requisite power outputs. Once efficiency was accounted for, using ground beef and a simple measurement circuit confirmed the performance of the commercial electrical neuromodulation devices. Beyond testing two devices, the development of rapidly prepared, low-cost phantoms, a method to measure the local activity of each cubic sub-volume, and a performance metric to allow for direct comparison, would enable the experimental streamlining of neuromodulation parameter optimization.

4.2. Limitations

Several limitations must be acknowledged in this study’s implementation. The primary limitation was the low efficiency (and thus, low sensitivity) of the recording system. While the electrodes were physically fixed, their positions constantly shifted. Only a small portion of the recording probe (a metal cone with a diameter of 0.7mm) was exposed to the material at a given time. Constant shifting between positions changed how much possible contact was measured between the immediate material and the electrode. Noise during recording caused rapid shifts, as did the slow sampling frequency of the system. Another problem was the continued deformation of the soft phantom materials, which shifted when recording electrode positions across repeated measurements. An issue with the relatively small number of electrodes is the limited resolution of electrical activity in each smaller volume. Another issue was the lack of comparison with high-resolution simulations or live trials. The current model was limited to electrodes, whereas other phantoms support multimodal sensors. However, these shortcomings are amenable to possible iterative improvements.

4.3. Future Work

Phantom brain fabrication for neuromodulation evaluation is an active field. The operational amplification gain could be adjusted to match those used in sampling weak electrophysiological signals, as in electroencephalography, to record efficiency issues [12,14]. A faster recording device could be used to improve the sampling frequency. Integrating the electrodes into the phantom’s container or mold, such as via sacrificial or 3D-printed scaffolding, could address electrode displacement in future iterations [10]. Compressing the phantom material before each test might also preclude shifting during testing. Other types of electrodes, including variance in geometry and materials, should be explored. Comparisons between electrode types, device configurations, measurement resolutions, and experiments could use EFs and observed dosages to determine which are the most practical in time and cost. Because the goal of this study was to reduce experimental complexity for rapid, iterative tests, high-resolution simulations would be better suited to more theoretical applications. Whereas the current phantom model relied on electrodes, future works could integrate other sensors, such as thermal or mechanical, to further understand the possible secondary effects of each electrical stimulation modality [13]. By deploying a phantom with readily available, low-cost materials, consumer neuromodulation device performance can be accurately and safely modeled. While different studies aim to test specific parameters, the use of consistent performance metrics can similarly enable a larger, faster, and more systematic comparison between stimulation parameters.

5. Conclusions

The variety of consumer neuromodulation devices increases annually, requiring an expedient and cost-effective means to verify manufacturer claims [7]. Because clinical trials and animal tests are costly and risky, using a brain phantom can facilitate preliminary experimental measurements [9]. For well-researched modalities such as tDCS and newer modalities such as am-tPCS, a range of materials was tested for use as electrical phantoms. Based on cost, preparation time, and efficiency, ground beef and agar gel with a 10% salt mix were selected. A method was proposed to measure local activity in each local volume of the phantom, without the need for complex simulations or specialized materials. The measured values for the total dosages were 0.55 W-s for am-tPCS and 0.91 W-s for tDCS. Due to a low gain on the recording electrodes, the signal efficiency measured against power delivered was 4.2% for tDCS and 3.1% for am-tPCS. The gain of the recording system and electrode positioning could be optimized in future iterations [13,16]. Thus, a simple recording system and low-cost ground beef phantom could facilitate the rapid testing, experimentation, and verification of consumer neuromodulation devices. EFs, as calculated here, could serve to further compare phantom studies in the future. Thus, the expedient, low-cost phantom, combined with a consistent performance metric, enables the rapid and direct testing and comparison of electrical stimulation parameters without requiring human or animal studies.